Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
Paper 12-0220
1
Abstract — This paper presents the design and experimental
test of the robotic elbow exoskeleton NEUROExos
(NEUROBOTICS Elbow Exoskeleton). The design of
NEUROExos was focused on three solutions which enable its use
for post-stroke physical rehabilitation. Firstly, double-shelled
links allow an ergonomic physical human-robot interface, and
consequently a comfortable interaction. Secondly, a 4-degree-of-
freedom passive mechanism, embedded in the link, allows the
user’s elbow and robot axes to be constantly aligned during
movement. The robot axis can passively rotate on the frontal and
horizontal planes 30° and 40° respectively, and translate on the
horizontal plane 30 millimeters. Finally, a variable impedance
antagonistic actuation system allows NEUROExos to be
controlled with two alternative strategies: independent control of
the joint position and stiffness, for robot-in-charge rehabilitation
mode, and near-zero impedance torque control, for patient-in-
charge rehabilitation mode. In robot-in-charge mode, the passive
joint stiffness can be changed in the range 24-56 N∙m/rad. In
patient-in-charge mode, NEUROExos output impedance ranges
from 1 N∙m/rad, for 0.3 Hz motion, to 10 N∙m/rad, for 3.2 Hz
motion.
Index Terms — Rehabilitation Robotics, Wearable Robotics,
Smart Actuators, Human-Robot Joint Axes Self-Alignment,
Physical Human-Robot Interaction.
I. INTRODUCTION
troke is the major cause of adult long-term disability in
Europe and many other countries [1]-[3] and strains
national services and related costs [4]-[6]. In about 85%
of cases, stroke causes hemiparesis in subjects, resulting in
impairment of the upper limb and disabilities in performing
activities of daily living, with consequent medical and social
care consuming a huge amount of healthcare resources [7].
Robot-aided physical rehabilitation has been proposed to
support the physicians in providing high-intensity therapy,
Manuscript received on April 29, 2012. This work was partly supported
by: EU within the NEUROBOTICS Integrated Project (FP6-IST-FET Project #2003-001917), the EVRYON Project (FP7/2007-2013 Grant Agreement
num. 231451) and the CYBERLEGs Project (FP7/2007-2013 Grant
Agreement num. 287894); and Regione Toscana under the Health Regional Research Programme 2009 within the project EARLYREHAB.
All authors are with The BioRobotics Institute, Scuola Superiore
Sant’Anna, viale Rinaldo Piaggio, 34, 56025, Pontedera (PI), Italy. Nicola Vitiello is corresponding author to provide phone: +39 050 883472; fax:
+39050 883 497; e-mail: [email protected]. Emanuele Cattin was with The
BioRobotics Institute (formerly ARTS Lab) of Scuola Superiore Sant’Anna, viale Rinaldo Piaggio, 34, 56025, Pontedera (PI), Italy, until 2009.
consisting of repetitive movements of the impaired limb [7]-
[9]. Robots can allow patients to receive a more effective and
stable rehabilitation process, and therapists to reduce their
workload. Robots can also offer reliable tools for functional
assessment of patient progress and recovery by measuring
physical parameters, such as speed, direction, and strength of
patient residual voluntary activity [10]. Robot-aided
rehabilitation is slowly convincing the community of
therapists to be as good as or even better than manual therapy
[10]-[12].
Common architectures of rehabilitation robots include: end-
point manipulators [11], [13]-[16], cable suspensions [17]-
[19], and powered exoskeletons [20]-[33]. Among these,
powered exoskeletons, despite their higher system complexity,
can provide assistance independently to each user’s joint. This
allows to better retrain the correct physiological muscle-
skeletal synergies, minimizing and controlling any
compensatory movement.
An exoskeleton for post-stroke physical rehabilitation is a
non-portable mechanical device that is anthropomorphic in
nature, is “worn” by the user and fits closely to his or her body
[33]. Given the close interaction with the user, comfort is a
major concern. The robot should be lightweight and take into
account the user’s joints range of motion (ROM),
anthropometry, and kinematics [32], [34]. The physical
human-robot interaction (pHRI) area should be large and
match the shape of the patient’s limb, to reduce the pressure
on the user’s skin [27], [35]. Furthermore, the actuation and
control of the robot should allow safe execution of
rehabilitation exercises in two modes of operation: robot-in-
charge, when the robot is driving the subject in doing the
NEUROExos: a powered elbow exoskeleton for
physical rehabilitation
Nicola Vitiello, Member, IEEE ̧Tommaso Lenzi, Student Member, IEEE, Stefano Roccella, Stefano Marco
Maria De Rossi, Student Member, IEEE, Emanuele Cattin, Francesco Giovacchini, Fabrizio Vecchi, Member,
IEEE, Maria Chiara Carrozza, Member, IEEE
S
Fig. 1 Overview of NEUROExos: (a) lateral view, (b) front view.
mailto:[email protected]
Paper 12-0220
2
Fig. 2 Double-shelled links: (a) exploded view of the links, (b) zoom
on the double-walled carbon-fiber structure of outer shells, and (c)
section view of the mechanism connecting outer and inner shells: (1)
layer of polypropylene, (2) layer of ethylene vinyl acetate (EVA), (3)
elastic bushing, (4) aluminium frame, (5) spherical joint, (6) threaded
rod.
exercises, and patient-in-charge, when the subject is driving
the robot that is only partially assisting the movement [11].
In this paper we introduce NEUROExos, a novel elbow
powered exoskeleton (see Fig. 1) designed for post-stroke
rehabilitation of the arm, ensuring maximum comfort and
safety to the patient. NEUROExos presents three innovative
design solutions:
1. a compact and light-weight mechanical structure with
double-shelled links, with a wide pHRI area to minimize
the pressure on the skin;
2. a 4-degree-of-freedom passive mechanism that unloads
the elbow articulation from undesired loads by ensuring
the alignment of human and robot joint axes;
3. an antagonistic, compliant remote actuation system with
an independent joint position and stiffness control (for
robot-in-charge exercises) and a near-zero impedance
torque control (for patient-in-charge exercises).
In the paper we also present the results of the experimental
characterization carried out to assess: the suitability of the
pHRI area for a comfortable interaction, the functionality of
the 4-DOF passive mechanism by comparing its degree of
laxity with the one of the human elbow axis, and the
performance of the actuation, sensory and control system.
NEUROExos was previously introduced by two conference
papers: in [36] we gave an overview of the design paradigm,
and in [37] we presented the passive compliance controller.
NEUROExos was also used as a test bed for control
algorithms and sensory system: in [38]-[40] we used it to test
two different algorithms for detecting user motor intentions,
and in [41] to validate a new sensing technology for
measuring human-robot interaction pressure onto a wide
tailored interaction surface.
The design of NEUROExos is described in Section II,
results of the experimental characterization are reported in
Section III and are discussed in Section IV. Finally, Section V
draws the conclusions.
II. THE NEUROEXOS PLATFORM
This section presents the main technical solutions of
NEUROExos. Five subsystems are implemented on the
platform and described hereafter: the double-shelled link
structure, the 4-DOF passive mechanism, the antagonistic
tendon-driven compliant actuation, and the control and
sensory systems.
A. Double-shelled links
Most upper-limb powered exoskeletons are made of bar-
shaped links, coupled with the user's limb segments through
multiple orthotic shells or cuffs [26]-[32]. This solution, while
simple, introduces problems in terms of encumbrance, inertia,
and kinematic compatibility with the limb, resulting in a poor
wearability of the robot. To overcome these limitations, the
links of NEUROExos are made of a double-shelled structure
(Fig. 2-a) composed of two concentric shells (inner and outer
shells).
Outer shells provide structural stiffness and strength to the
robot, and transfer the load to the human limb segments.
Compared with bar links, shell-shaped links have their center
of mass closer to the longitudinal axis of the human limb
segment thanks to an optimized material distribution around
the limb segment. This reduces the possibility that the
assistive load applies an uncomfortable torque about the
longitudinal axis of the human limb segment.
Inner shells are in direct contact with the user’s arm to
transfer the loads. Thanks to the use of a soft orthopaedic
material and a wide interaction area, they contribute to reduce
the pressure on the user’s skin and ensure a comfortable
interaction [41].
1) Inner shells Each inner shell is made of two half-shells, coupled with the
dorsal and ventral sides of the arm (Fig. 2-a) and fastened with
velcro belts, as shown in Fig. 1. Inner shells have a two-
layered structure: a 3 mm-thick internal layer of ethylene vinyl
acetate (555XEB/3, M.T.O., Italy), for moisture draining and
skin transpiration, and a 3 mm-thick outer layer of
polypropylene (558/3 M.T.O., Italy). Inner shells come in
Paper 12-0220
3
Fig. 3 Anatomy of the human elbow: (1) humerus, (2) radius, (3)
ulna, (4) capitellum, (5) throclea, (6) lateral facet of capitellum, (7)
lateral facet of trochlea. AH is the humerus longitudinal axis, AU is
the ulna longitudinal axis, AML is the anatomical medial-lateral axis
passing from the capitellum center to the trochlear center [46], βh and
βf are the frustum vertex angles respectively on the horizontal and
frontal planes (adapted from [71]).
different sizes, and can also be tailor-made on each subject
(e.g. by thermo-shaping a polypropylene layer).
2) Outer shells Outer shells have a double-walled carbon fibre structure,
which has total height of 10 millimeters and thickness of 1.5
millimeters (see Fig. 2-b). Size and shape of the NEUROExos
outer shells were designed by using a 3D model of the human
arm surface. This surface was obtained by laser-scanning
(INKAY, Italy) and digitalization (Pro/Engineer, PTC, MA,
USA) of the arm surface of a voluteer (male, height 175 cm,
weight 75 kg). Outer shells can be connected to inner shells of
different shapes, and, therefore, allow the same exoskeleton
links to be used by several users. Outer shells also house the
aluminium frames of the 4-DOF passive mechanism, the gear
box for the active flexion-extension DOF, and the inner-outer
shell connecting elements (Fig. 2-a).
NEUROExos upper-arm and forearm links (including all
actuation and mechanics components) have a total weight of
1.65 kg and 0.65 kg respectively, and the moment of inertia of
the forearm link about the flexion-extension axis is equal to
7.2·10-3
kg·m2.
3) Inner-outer shell connecting elements The connection between inner and outer shells is obtained
through a small mechanism that allows small relative
adjustments.
The connection is depicted in Fig. 2-c. An aluminium frame
is embedded within the carbon fiber structure and houses a
ball joint (GE 8C, SKF, Sweden). A threaded rod passes
through the ball joint and is screwed into an elastic bushing
(Radialflex M4, Paulstra, France), which is connected to the
inner shell. Turning the threaded rod allows the regulation of
the inter-shell distance (dimension d in Fig. 2-c) in a range of
15 millimiters. Four connecting mechanisms are used for each
couple of inner-outer shells. The relative spatial orientation
between inner and outer shells is set by independently
changing the distance d at each connection point. Passive ball
joints allow the rods to passively tilt of a maximum angle α of
15° (see Fig. 2-c), thus preventing the inner shells from being
strained. Elastic bushings provide the connecting mechanisms
with a high longitudinal (along the threaded rod longitudinal
axis) compression stiffness (100 N/mm), which is needed for
load transferring between inner and outer shells, and a low
tangential stiffness (15 N/mm) allowing small relative sliding
motions between shells.
B. The 4-DOF passive mechanism
Exoskeletal machines are worn by the user, therefore they
should match the constraints given by the kinematics of the
limb to be assisted because misalignements between human
and robot joint rotation axes can cause translational forces at
the pHRI surface [43]. These translational forces are highly
undesired, since they load the skin and the muscleskeletal
system, and make the interaction uncomfortable or even
painful [34].
The problem of joint axes misalignment is particularly
critical in exoskeletons interacting with the upper limb in
multiple points (i.e. hand, forearm, upper-arm, and/or trunk).
Given that each limb segment is connected rigidly to the robot,
translational forces are entirely unloaded on the user’s skin
and articulation. A correct alignment is difficult to achieve in
exoskeletons, given that the exact location and orientation of
human joint axes cannot be detected from the outside without
complex imaging techniques. Moreover, many human joints
do not behave as simple hinges, changing the spatial
configuration of their rotation axis along with the joint motion.
The misalignment problem has been often addressed by
designing specific kinematics schemes [26], [28]-[31] or joints
[27], [32]. However, a more general methodology to address
this issue has been proposed in [34]. The basic idea consists of
mounting the active rotational joint on a moveable
translational passive mechanism which decouples the robot
joint rotations from axis translations. Translational passive
mechanisms unload human limb segments and articulations
from undesired translational forces, while the rotational active
joints transfer the assistive torques onto the human joints [34].
This solution, however, does not take into account the laxity
of the elbow articulation. Elbow laxity has been widely
investigated in orthopedics studies, showing that elbow
articulation behaves as a “loose hinge joint” [44]-[46]. Over
the flexion-extention motion range, the elbow rotation axis is
not firm, but traces the surface of a double quasi-conic frustum
with an elliptical cross-section [44]-[45] (see Fig. 3). The
double frustum has both an inter- and intra-individual
variability, the latter being determined by: the flexion mode
(active or passive motion of the joint), the forearm position
(pronated or supinated) and any varus or valgus torque loading
Paper 12-0220
4
Fig. 5 Description of the passive movements obtained by means of
the NEUROExos 4-DOF passive mechanism: (a) rotation in the
frontal plane, (b) rotation in the horizontal plane, (c) translation of the
forearm link along the flexion-extension axis, and (d) translation in
the horizontal plane.
Fig. 4 Schematic drawings and pictures describing the strucutre and
working principles of the NEUROExos 4-DOF passive mechanism:
(a) CAD and (b) layout of the passive mechanism: (1) NEUROExos
flexion-extension joint, axis AFE, (2) prismatic joint trough splined
shaft, (3) universal joint, (4) circular slider, (5) carbon fiber link, (6)
linear slider, (7) spherical joint, (8) spherical joint, (9) rotational
joint, (c) zoom on joins (1), (2), (3) and (4): AUv and AUh are the
vertical and horizontal rotational axes of the universal joint, the
splined shaft allows for the prismatic joint (2) whose axis coincides
with AFE, (d) implementation of the passive mechanism.
the articulation. In particular, the frustum vertex angles βf, on
the frontal plane, and βh, on the horizontal plane, assume a
maximum value of about 10° and 6° respectively. In addition,
the elbow average rotation axis over a full flexion-extension
task forms an angle of 80°-92° with the humerus longitudinal
axis AH onto the frontal plane, and an angle of ±5° with the
medial-lateral anatomic axis AML onto the horizontal plane.
Starting from the variability of the human elbow laxity, we
designed a 4-DOF passive mechanism which provides the
NEUROExos powered axis with the same degree of laxity of
the human elbow. Thanks to this passive mechanism, the
active axis can trace a double conoid whose frustum vertex
angles satisfy both the intra- and inter-subject variability of
elbow axis laxity. In addition, two passive translational DOFs
unload the elbow articulation from undesired translational
forces.
The passive mechanism scheme along with its
implementation is depicted in Fig. 4. The mechanism consists
of a closed-chain composed of 13 passive joints: 4 prismatic, 4
spherical, 2 circular sliders, 2 universal and 1 rotational joint.
Despite its complexity, this compact mechanism fits within the
space between upper-arm inner and outer shells, and therefore
does not affect the overall system encumbrance.
NEUROExos flexion-extension axis AFE is identified by the
two axes of the prismatic joints labeled as 2 in Fig. 4. These
joints are implemented by means of two splined shaft-hole
couplings having a ROM of 35 millimeters. Each splined shaft
is attached to a universal joint (labeled as 3, see Fig. 4), whose
ROM is 100° around its horizontal axis (AUh) and 24° around
its vertical axis (AUv) (see Fig. 4-c). Each fork housing a
universal joint is then attached to a slider (joint number 4 in
Fig. 4) which moves along a circular trajectory having a
diameter of 120 millimeters and an angular ROM of 42°.
Paper 12-0220
5
Fig. 6 Schematic drawing of the antagonistic tendon-driven
compliant actuation and sensory system of the NEUROExos: (a) two
remote antagonistic units, named flexor (‘flx’) and extensor (‘ext’),
power the NEUROExos active joint, (b) exploded view of the driving
block, and (c) working principle of the cable force sensors.
Through an L-shaped carbon fiber link, the circular slider is
connected to a linear slider (joint number 6, ROM of 30
millimeters). The linear slider is linked to the rotational joint
labeled as 9 (ROM of 40°) by means of two spherical joints
(male threaded, maintenance-free rod ends, SKF, Göteborg,
Sweden), labeled as 7 and 8, connected by a bar with an
adjustable length.
The passive mechanism provides four DOFs as depicted in
Fig. 5. Three DOFs are used to allow the NEUROExos axis
AFE to trace a double conic frustum:
1. AFE can rotate in the frontal plane of an angle γf = ±15°
(see Fig. 5-a);
2. AFE can rotate in the horizontal plane of an angle γh =
±21° (see Fig. 5-b);
3. the NEUROExos forearm link can slide along the axis
AFE of a distance of ±15 mm (see Fig. 5-c).
This latter DOF, which is determined by the prismatic joints
labeled as 2 in Fig. 4, allows the self-alignment of the waist of
the double quasi-conic frustum traced by the elbow axis with
the vertex of the double conoid traced by the NEUROExos
axis along the anatomical medial-lateral axis AML.
The fourth DOF, shown in Fig. 5-d, allows the AFE to
translate on the horizontal plane along the antero-posterior
direction of a segment Δh= ±15 mm. Therefore it allows any
undesired translational force acting on this direction to be
unloaded [34].
Finally, the elastic bushings of NEUROExos allow the
user’s upper arm (rigidly connected to the upper inner shell) to
slide against the NEUROExos upper shell. This way an
additional DOF is given to the NEUROExos axis AFE which
unloads the subject’s articulation from the frontal-plane
component of the undesired translational force.
C. Antagonistic compliant actuation
The actuation and control system of an exoskeleton for post-
stroke physical rehabilitation should provide two different
therapy protocols. The two modalities can be described as
robot-in-charge and patient-in-charge [47], [48]. In robot-in-
charge mode, which is usually implemented in the initial stage
of the rehabilitation process when patients cannot move
autonomously, the robot should be able to promote a desired
motion pattern. This modality requires the robot to have
relatively high joint impedance. Patient-in-charge mode is
needed when the subject can control some movements of
his/her limb, and requires some assistance from the system to
complete the task. In this latter case, the robot should not
hinder the patient motion, but rather it should show near-zero
impedance while assisting the user. Therefore it is
fundamental to provide the robot with impedance control.
Active control of impedance can be achieved using rigid
actuation systems, such as geared electric motors, and a
closed-loop torque control. Unfortunately, beyond the closed-
loop control bandwidth, the system will show high inherent
impedance [27]-[28], which raises strong safety issues [49].
Series elastic actuators have been successfully applied in this
field to solve safety issues [47], [48], [50], [51]. In this case,
the actuation is not rigid and allows relatively low joint
impedance across the entire frequency spectrum. However,
variations in the output impedance can still be achieved only
by means of closed-loop interaction control strategies [49].
The so called “actively-adjustable passive compliance”
actuators [52]-[55] overcome these limitations. This kind of
actuation system can provide software-controllable hardware
compliance, simplifying the control system, minimizing the
risk of instabilities of active impedance control, and
Paper 12-0220
6
guaranteeing maximum safety in the interaction with the
robot. The adjustable compliant behavior of this actuator
system is obtained by means of an inherent hardware property
of the actuation, requiring no closed-loop control.
NEUROExos is powered by an antagonistic-driven
compliant joint (ADCJ), an actively-adjustable passive
compliance actuator successfully exploited in [20], [25]. The
ADCJ is powered by two antagonist actuation units, each
showing a non-linear elastic behaviour with an adjustable
resting position.
The layout of the ADCJ implemented on NEUROExos is
depicted in Fig. 6-a. It consists of a pair of remote and
independent antagonistic units [56]-[60]. Each unit consists of
a series of: a non-linear elastic mechanism with a quadratic
force ( ) vs. cable elongation ( ) function (
with N/mm2 and N/mm), a
linear hydraulic actuator with a stroke of 50 millimeters
(Parker-Hannifin Corp., OH, USA), a stroke amplifier
(transforming a hydraulic piston displacement into an elongation of the transmission , where is the amplification ratio), and a steel-wire rope with a diameter of
1.4 millimeters (Carl Stahl, Süssen, Germany) which transmits
the force to the NEUROExos driving block through a Bowden
cable.
The non-linear elastic mechanism is based on a linear
tension spring coupled with a cam mechanism (see Fig. 6-a),
which was presented and characterized in [57]. By using a
tension spring of 80 N/mm (T series, D.I.M. srl, Italy)
NEUROExos achieves a passive joint stiffness in the range
20-60 N m/rad. This range is comparable to the one of the human elbow measured in single- [61] and multi-joint arm
movements [62] and thus prevents the subject from an
uncomfortable (or even painful) interaction with an
excessively stiff device in case of involuntary spastic
movements. Moreover, the stiffness range of NEUROExos is
also comparable to that generated by state-of-the-art end-point
manipulators on the elbow [11].
Each hydraulic cylinder is controlled by a three-land-four-
way proportional electronic valve, commanded by a ±10 V
DC signal, which sets the piston velocity [56], [60]. The
hydraulic circuit is powered by a three-phase 1.1 kW AC
motor (Parker-Hannifin Corp., OH, USA).
Fig. 6-b shows an exploded view of the driving block, which
has a maximum output torque of 15 Nm. This unit is
composed by the driving pulley (radius of 19 millimeters),
which the antagonistic steel ropes wrap around, a custom-
made planetary gear that amplifies the torque by a factor of
four, the frame housing the force sensors, and two mechanical
end stops that prevent elbow hyperextension and hyperflexion.
Although the planetary gear slightly increases the
encumbrance and the inertia of the driving block, its use
enables the force transmitted by each antagonist cable to be
lowered by a factor of four. This is beneficial for two reasons.
First, the lower the trasmitted force, the lower is the friction
loss given by the Bowden cables, as outlined in [63] and
discussed in our previous works [56]-[60], [64]. In fact, high
values of friction would affect the static passive behaviour of
the NEUROExos joint when controlled in passive-compliance
control mode (see Section E.1): higher the friction, higher the
discrepancy between the desired and actual torque field.
Second, lowering the transmitted force on the tendon cable
also reduces - by the same factor four - the full-scale range of
the cable force sensor, which can therefore have smaller size
and, consequenlty, the overall design of the driving block can
be more compact.
D. Sensory apparatus
A 1024 ppr incremental optical encoder (2420, Kübler,
Germany) was assembled coaxially with the driving pulley
(see Fig. 6-a and Fig. 6-b) and the sun gear to measure the
flexion-extension rotation angle (resolution of 0.022°).
Two custom-made load cells were included in the design to
measure the force transmitted by the antagonist tendon cables.
In Fig. 6-c the working principle and the layout of the force
sensors is depicted. Each antagonist cable comes out from the
Bowden cable, passes through an idle pulley, and then wraps
on the driving one. The idle pulley is hinged on a shaft rigidly
connected to the cantilever of the force sensor and deflects the
cable of an angle φ of 15°. Consequently, a component of the
cable force ( )) bends the cantilever. The strain of the cantilever, which is linearly dependant on , is measured by means of four piezoresistive strain gauges (ESU-025-1000,
Entran, England, UK) mounted in full-bridge configuration,
and conditioned by a commercial electronics (MecoStrain,
MECO, Italy). The mechanical structure of the force sensor
was designed to work in safe condition against any overload.
In fact, a mechanical end stop prevents the cantilever from
becoming loaded over the yield stress and limits the maximum
sensed cable force to 200 N. Voltage-to-force curves of the
force sensors were characterized and exhibited high linearity
(RMSE=0.05 N and R2=1) and low hysteresis (0.05% of the
full-scale range). Peak-to-peak noise was quantified as 0.05 N,
constant over the full-scale range.
Two linear potentiometers (SLS095, Penny&Giles, Dorset,
UK) are used for the measurement of the piston positions with
an accuracy of 0.01 millimeters.
E. Control system
The actuation system of NEUROExos allows for the use of
two alternative control strategies: the passive-compliance
control and the torque control, respectively for the execution
of robot-in-charge and patient-in-charge exercises.
1) Passive-compliance control NEUROExos antagonistic actuation and passive-
compliance control take inspiration from the human
musculoskeletal system, which powers the limbs by using
antagonistic muscle pairs. The musculoskeletal system
generates a convergent force field around an equilibrium
position of the limb by relying on the elastic properties of
antagonistic muscles. The selective activation of one of the
two muscles displaces the convergent field towards a new
equilibrium position and, consequently, changes the position
of the limb. The simultaneous co-activation of both muscles
(i.e. muscles co-contraction) increases the slope of the
convergent field (i.e. the joint stiffness), leaving the limb
Paper 12-0220
7
Fig. 7 Block diagram of NEUROExos control strategies: (a) passive-
compliance control, (b) torque control.
position unchanged [65]-[69]. In a similar way, the actuation
system of NEUROExos can apply a convergent torque field
around a certain angular position (i.e. the equilibrium point)
by regulating the rest lengths of two opposite elastic actuation
lines. The slope of this convergent torque field (i.e. the joint
stiffness) can be regulated independently of the equilibrium
position, thanks to the non-linearity of the compliant elements
[69].
The torque applied by the antagonistic cables on the NEUROExos active joint is:
(1)
where is the radius of the driving pulley, is the
transmission ratio of the planetary gear, and and are
the force applied by the flexor and extensor units respectively.
Assuming the steel cable is infinitely stiff, the total elongation
of the transmission line coincides with the elongation of the non-linear elastic element and, consequently, the force driven
by each cable is a non-linear function of the spring elongation:
. (2)
The elongation , of each actuation unit, depends linearly on both the piston position and the joint angle
(3)
where and are the piston positions of the extensor
and flexor units respectively, is a fixed reference angle, and
and
are the positions for which the elongations
and are nil when ( , and
corresponds to the configuration of maximum extension). The joint equilibrium position is easily calculated by
making , and substituting (3) in (2):
(4)
where and
. By
appropriately changing the reference frames for the piston
positions, it is possible to have
, so that (4)
becomes:
. (5)
The joint stiffness, defined as , is equal to:
(6)
As equations (5) and (6) show, the joint equilibrium position
is proportional to while the joint stiffness
changes linearly with . Thereby, the joint
equilibrium position and stiffness can be regulated
independently.
NEUROExos joint passive compliance is controlled by
means of a two-layer hierarchical control system.
The high-level layer is dedicated to the coordination of the
piston positions. Given the desired passive compliance, i.e.
desired joint equilibrium position ( ) and stiffness (
),
the high-level layer calculates the desired piston positions of
both the flexor and extensor units by using two new control
variables, which are a linear combination of and (see
Fig. 7-a). The differential-mode command, is defined as a
reciprocal shift of the antagonist pistons:
. (7)
The common-mode command, is an equal shift of the
antagonist pistons:
(8)
Substituting (7) and (8) into (5) and (6), we get the final
equations describing how differential- and common-mode
commands can be used to respectively control the joint
position (9) and stiffness (10):
(9)
. (10)
The differential- and common-mode commands are then
converted into piston positions:
. (11)
The low-level layer controls the hydraulic piston positions
and by means of two independent PI closed-loop
regulators [57] with a 20 Hz bandwidth.
In a robot-in-charge task, the passive-compliance controller
moves the human elbow by displacing the equilibrium
position of NEUROExos along a desired trajectory. The
passive-compliance controller acts in open-loop fashion with
respect to , and generates a torque field proportional to the
and . This way, NEUROExos does not force
the position of the human joint (like a position servo would
do), but rather it allows deviations from the predefined path:
the actual position is regulated by the following dynamics equation:
(12)
where [N·m·s2/rad], [kg] and [m] denote the inertia, mass and equivalent length of human forearm and hand
coupled with the NEUROExos forearm module, [m/s2] denotes the constant of gravity, and denote respectively the torque applied by the NEUROExos actuation,
i.e. , and the torque applied by
Paper 12-0220
8
Fig. 8 Plots of experimental data recorded during the characterization
of 4-DOF passive mechanism: (a) reference coordinate system and
instantaneous rotation axes AFE(n) for Subject 1, (b) , Δh(t), γh(t) and γf(t) for Subject 1.
human muscles activation on the elbow, takes into account the friction loss in the planetary gear and the Bowden
cables, and is the angle between the longitudinal axis of
the upper-arm link and the gravity vector (adjustable in the
range [0°,45°] by an external frame).
In quasi-stationary (i.e. ) condition and ignoring static friction losses, the actual position depends solely on the human muscles activation and the gravity:
(13)
In this case, it follows that when
, which happens when the human voluntary
action balances the gravity action.
2) Torque control In order to be used for patient-in-charge control strategies,
torque control should be able to provide the patient with an
assisitve torque with near-zero output impedance, i.e. with
minimum to null joint parasitic stiffness.
As shown in the block diagram of Fig. 7-b, the torque
control of NEUROExos relies on the independent closed-loop
control of the cable force powered by each actuation unit. The
desired torque is converted to desired forces on the antagonistic cables by means of the following equation:
(14)
where is a preload force constantly
applied to both the antagonist cables and
. Then, the desired cable forces serve as
input of two independent closed-loop force controllers. The
closed-loop control architecture is that of a classical PID
regulator, with a saturation interval of [-0.14, 0.14] m/s for the
speed of the hydraulic piston, and an anti-wind-up scheme.
The PID regulator operates on the error between the desired
and measured cable forces and outputs the speed of the
hydraulic piston, which is controlled by means of the DC
proportional electro valve (see Section III.C). PID regulators
were tuned manually for achieving the widest possible closed-
loop bandwidth.
By setting the preload force we tune the physical joint stiffness, which is then lowered by the action of the closed-
loop controllers. The relationship between preload force and
physical stiffness can be obtained by reversing (2)-(3) and
applying (6):
(15)
3) Control unit and safety loop NEUROExos controllers run on a real-time control system
(PXI-8196 RT, National Instrument, Austin, TX, USA)
equipped with a data acquisition card (M-series, National
Instrument, Austin, TX, USA). The high-level layers run at
100 Hz, while the low-level closed-loop (position and force)
controllers run at 1 kHz. Signals of both cable force and piston
positions sensors are sampled at 250 kHz, then low-pass
filtered and down-sampled to 1 kHz.
The NEUROExos control system implements a safety loop
that switches off the actuation when the force on a cable
exceeds 150 N, the joint torque 10 Nm, or the joint speed is
greater than 400 deg/s. In addition, in order to detect possible
failures of the force sensors and prevent the user from injuries
and the system from damages, the safety loop compares the
output of the force sensor with an estimate of obtained through (2) and (3), and switches off the actuation when the
difference is more than 30 N.
III. EXPERIMENTAL CHARACTERIZATION
In this section, we characterized the 4-DOF passive
mechanism (Section II.B), by testing its effectiveness in
aligning the robot with the user’s elbow axis, and the
performance of the two control strategies (Section II.E).
A. Characterization of the 4-DOF passive mechanism
Five healthy subjects (3 males and 2 females) volunteered
to participate in the experiment. Each subject wore
NEUROExos and performed a cyclical flexion-extension
movement (amplitude of about 100°, frequency of about 0.35
Hz, total duration of 120 s). Actuation was unplugged during
the experiment. The AFE rotation axis was tracked by means of
an optical motion capture system (460, VICON, Oxford, UK)
using six passive optical markers. Two markers were placed
Paper 12-0220
9
on upper- and forearm external shells to identify the
longitudinal axis of each link. Two markers were applied
coaxially to the driving pulley to identify the AFE axis.
The motion of the AFE rotation axis was described in terms
of the two angles γf and γh, which indicate respectively the
rotation on the frontal and on the horizontal planes, and a
translation Δh on the horizontal plane. For each subject and
each trial, we applied the following four-step algorithm to
identify the horizontal and frontal planes of the human elbow
articulation and to compute γf, γh and Δh.
Step 1) For each time sample n between 1 and N, we
extracted a geometrical representation of the rotation axis
AFE(n), and of the longitudinal segment of the NEUROExos
upper-arm UA(n) and forearm FA(n) links (UA(n) is fixed).
Step 2) We calculated the average rotation axis av
AFE over
the entire set of recorded rotation axes AFE(1 … N).
Step 3) We identified three orthogonal datum planes.
Among the infinite planes orthogonal to av
AFE, the reference
Sagittal Plane (rSP) was chosen to minimize the average
distance between the intersection points of AFE(1 … N) and
rSP with av
AFE. The reference Frontal Plane (rFP) was
defined as the plane orthogonal to rSP which passes trough av
AFE and the segment UA. Finally, a reference Horizontal
Plane (rHP) was defined to be orthogonal to both rSP and rFP,
and passing through av
AFE. In Fig. 8-a, the three datum planes
along with AFE(1 … N) are depicted for Subject 1.
Step 4) We computed γf(n) as the angle between rFP and the
projection of AFE(n) on rSP. Likewise, γh(n) was calculated as
the angle between rHP and the projection of AFE(n) on rSP.
Δh(n) is obtained from the projection of the distance between
AFE(n) and av
AFE onto rHP.
Fig. 8-b shows the position of AFE for Subject 1, in terms of
γh, γf and Δh, along with the elbow flexion-extension angle .
Table I reports the mean absolute value, the minimum and
maximum of γh, γf and Δh for all subjects, plus the average of
these values over all subjects.
B. Characterization of the passive-compliance control
In this Section we evaluate the static and dynamic
performances of the passive-compliance control. All the
experiments were performed with .
1) Static characterization The static characterization aims to verify the NEUROExos
passive joint stiffness performances in static conditions (i.e.
). The equilibrium position was set to , then,
for five different common-mode commands (i.e. ), we manually displaced the NEUROExos joint about 15° in both directions in quasi-static conditions
(i.e. ). For each value of the procedure was iterated ten times in both flexion and extension directions.
The results of this characterization are shown in Fig. 9-a.
The measured torque increases linearly with the absolute value
of the difference between the equilibrium position and the
actual position . An increase of the common-
mode command results in a higher slope of the torque vs.
angular displacement curve, and therefore, in an increased
passive joint stiffness. The joint stiffness values were
estimated through linear fitting (see Fig. 9-a) and are reported
in Table II.
2) Dynamic characterization To characterize the dynamic behaviour of the passive-
compliance control, we performed a step and chirp response
analysis with three different common-mode values. The
passive compliance control was also tested in a prototypical
robot-in-charge task.
Step response) To characterize the adjustable dynamic
behaviour of the variable-compliance joint, the step and chirp
response analysis was executed with NEUROExos unloaded,
i.e. neither did a subject wear it nor were additionally mock-up
masses connected to the moving link. A 30° position step was
given at three levels: 0, 1 and 2 mm. Twenty repetitions were performed. Fig. 9-b shows the angular trajectory
averaged over all the iterations for each stiffness level. As
reported in Table III, the rise time decreases proportionally
with the increase of the joint stiffness. Moreover, the steady-
state angular error decreases with the level of stiffness (see Table III).
Table I Results of the 4-DOF passive mechanism characterization.
For each subject, we reported the mean of the absolute value, the
maximum and the minimum of γf [deg], γh [deg] and Δh [mm], along
with their averaged value overall five subjects.
Subject #1 #2 #3 #4 #5 Mean ± std
mean(|γf|) 2.38 2.08 2.05 3.46 2.45 2.49±0.56
max(γf) 6.13 6.05 7.08 8.68 8.96 7.38±1.29
min(γf) -9.62 -6.81 -5.97 -14.2 -9.03 -9.12±1.05
mean(|γh|) 1.90 1.88 1.95 1.73 0.61 1.62±0.56
max(γh) 3.22 4.61 5.46 3.05 2.25 3.72±1.29
min(γh) -4.40 -4.13 -4.13 -4.96 -2.16 -3.96±1.05
mean(|Δh|) 0.52 0.84 1.79 1.37 0.86 1.07±0.52
max(Δh) 2.89 5.56 9.52 12.5 6.39 7.39±3.73
min(Δh) -2.27 -3.37 -5.12 -5.55 -8.82 -5.03±2.49
Table II Static Characterization: fitting results.
[mm] 0 1 2 3 4
[N·m/rad] 24.6 29.2 36.6 46.4 56.7
RMSE [N·m/rad] 0.06 0.04 0.07 0.06 0.13
Table III Angular step response results.
[mm] 0 1 2
Rise Time [s]
0.071 0.067 0.064
Steady-state [deg] 0.51 0.23 0.12
Table IV Angular chirp response results.
[mm] 0 1 2
-3dB bandwidth [Hz]
6.45 6.91 7.24
-3 dB phase [deg]
93.1 89.5 85.8
Table V Mean and the standard deviation of the amplitude difference
between the reference and actual angular, and RMSE of
, for the four joint stiffness levels.
[mm] 0 1 2 3
Amplitude
difference [deg] 22.2±1.4 15.6±2.2 6.1±1.8 2.2±0.4
RMSE of [deg] 12.1 8.06 5.7 4.96
Paper 12-0220
10
The step response is underdamped, with an overshoot of
3.7-19% and a steady-state error of 0.4-1.7% of the amplitude.
By increasing the joint stiffness, over-shoot, steady-state error
and rise time decrease of 80.8% (from 5.79 to 1.11 deg),
76.4% (from 0.51 to 0.12 deg) and 9.85% (from 0.071 to
0.064 s) respectively, and the peak velocity increases of
11.1%, from 420.1 to 467.2 deg/s.
Steady-state results from the combined action of the gravity and friction. In fact, the value of resulting from the sole action of the gravity (which can be computed by means of
(13) and assuming =0.65 kg and =10.36·10-2 m) is lower than the actual one of 0.66°, 0.75° and 0.66°, for equal to 1, 2 and 3 mm respectively.
Chirp response) The frequency response of the position
control was characterized by displacing the equilibrium
position along a linear chirp (frequency 0-8 Hz, duration 480
s, amplitude 30°). The same chirp was repeated for three
stiffness levels ( equal to 0, 1 and 2 mm). The estimated Bode diagram (amplitude and phase) of the system
was obtained as the ratio between the power
spectral density of the measured and input positions. Fig. 9-c
shows the resulting Bode plot for each stiffness level and
Table IV reports the -3dB bandwidth.
Prototypical robot-in-charge task) In order to evaluate the
functionality of the NEUROExos system, a prototypical task
was designed and tested on a healthy volunteer (male, 27 years
old, 70 Kg). This task simulates a simple rehabilitation
procedure, with the subject totally passive and the exoskeleton
driving his arm. NEUROExos was programmed to move the
equilibrium position along a sinusoidal trajectory (amplitude
105° deg, from 10° to 115°, frequency 0.5 Hz). A two-minute
sequence was repeated with four levels of joint stiffness,
obtained by setting equal to 0, 1, 2, 3 mm. Fig. 9-d shows the commanded and measured angular trajectories for
the four levels of stiffness. For the sake of clarity, only one
sine wave period is shown. It can be seen that there is an
angular difference between the equilibrium and the actual
trajectory. By increasing the joint stiffness, the difference is reduced. Table V reports the mean and standard deviation of
the amplitude difference between the equilibrium and the
actual angular trajectory, and the RMSE of over the entire duration of the sine wave, for the four levels of joint stiffness.
C. Characterization of the torque control
In order to characterize the closed-loop torque control
performance, we analyzed the response of the system to a
torque step and a torque chirp command, calculating also the
resulting output impedance.
1) Step and chirp response Both step and chirp responses were evaluated in static
conditions (i.e. ) with the NEUROExos joint being mechanically blocked and the preloading force set to .
The step response (from 1 to 7 N·m) was evaluated over 20
iterations. The averaged response is shown in Fig. 10-a. The
average value of the rise time was 0.054±0.002 s, the settling
time was 0.08±0.003 and the maximum overshoot was
0.27±0.007 N·m.
The chirp response was iterated three times. The reference
torque was a chirp signal with mean amplitude of 2 N·m (1 to
Fig. 9 Experimental characterization of the passive-compliance control: (a) joint torque vs. joint angle displacement curves, (b) step response
averaged over 20 iterations, (c) chirp response, (d) sine-wave prototypical task.
Paper 12-0220
11
3 N·m) and a 0-12 Hz linear frequency sweep over 600 s. The
resulting amplitude Bode diagram of the chirp response is
reported in Fig. 10-b, and the estimated -3 dB bandwidth was
10.35 Hz.
2) Characterization of the joint output impedance The output impedance of NEUROExos in torque control
mode was tested to assess quantitatively the effort that users
need to move the robot in zero torque mode (i.e. ). Impedance was evaluated by moving the joint in zero torque
mode and measuring the interaction with force sensors. The
transfer function from joint angle to actuator torque is an
estimate of the output impedance of NEUROExos in torque
control mode [48].
A volunteer wore NEUROExos and performed a quasi-
sinusoidal flexion-extension motion, with a torque reference
of zero and . The amplitude of the movement was about 30°, the frequency varied linearly in the range 0.3-3.2
Hz for total 40 s of recording. The movement pace was
indicated to the user by visual feedback and a metronome.
Five iterations were performed for statistical purposes.
Fig. 10-c shows the profile of the interaction torque felt by
the subject during the task, along with the profile of the
flexion-extension angle. It can been seen that the interaction
torque amplitude increases with the motion frequency,
reaching 1.80 N·m for 3 Hz motion.
Fig. 10-d shows the Bode plot of the transfer function from
the joint angle to the interaction torque. It can be seen that the
joint output impedance increases across the spectrum,
increasing from about 1 N·m/rad, for a 0.3 Hz motion, up to
about 10 N·m/rad at 3.2 Hz.
IV. DISCUSSION
This paper introduced the design of the robotic elbow
exoskeleton NEUROExos and presented the results of the
experimental activities which aimed at assessing: 1) the
NEUROExos pHRI surface, 2) the functionality of the 4-DOF
passive mechanism for aligning the robot and user’s flexion-
extension axes; 3) the performance of the antagonistic
actuation, and control system.
A. pHRI surface
The double-shelled structure allowed users with different
arm sizes to fit easily in the exoskeleton, thanks to the
different sizes of inner shells and to the laxity between inner
and outer parts.
Shaped inner shells maximized the human-robot contact area
reducing the pressure on the user’s skin and improving
comfort. This latter aspect was quantitatively assessed in [41],
showing typical peak pressures of 7.5 kPa, which is well under
the threshold of pain [72], [73]. In addition, despite inner
shells encompass the arm, they allowed unconstrained elbow
movement thanks to the compliance of the inner layer and of
the sylicon belts, which absorbed the volumetric changes of
upper-arm and forearm, caused by the muscular activity.
NEUROExos links (outer shells) provided sufficient
structural rigidity to transfer torques up to ±15 Nm, despite
the low weight (1.65 kg and 0.65 kg for the upper and lower
arm, resepctively), thanks to the double-walled carbon-fiber
Fig. 10 Experimental characterization of the NEUROExos torque control: (a) step response averaged over 20 iterations, (b) chirp response:
amplitude Bode diagram of the transfer function from desired to measured joint torque, (c) characterization of the joint output impedance:
angular displacement and interaction torque over a 3 Hz motion range, (d) Bode diagram of the transfer function from angular displacement to
interaction torque.
Paper 12-0220
12
structure.
B. 4-DOF mechanism
The 4-DOF passive mechanism aligns the human and robot
rotation axes, and follows the physiological displacement of
the elbow axis during flexion-extension movements (see Fig.
3). This mechanism (see Fig. 5), allows a frontal plane rotation
γf = ±15° and a horizontal plane rotation γh = ±21°. Such
ROMs are larger than the displacement of human elbow axis
during flexion-extension movements. In fact, human studies
showed that the observed frontal plane rotation βf was ±5° and
the horizontal plane rotation βh was ±3°. Having larger ROM
the 4-DOF mechanism can compensate for inter-subject
variability [44]-[46], which acts as an offset on the
physiological rotation ranges.
Results on five healthy subjects clearly showed that the
system can track the elbow axis on the whole movement
range. Fig. 8-b shows (for Subject 1) that the rotation angles γf
and γh have a periodic trend along with the task of flexion-
extension. Data in Table I, show that the average measured
ROM of γf and γh (respectively equal to 16° and 7°) complies
with the ROM of the human elbow axis found in [44]-[46].
Table I also shows that ∆h has an average measured peak-to-
peak ROM of about 10 millimeters. This shows that the
rotation axis moves on the horizontal plane during the flexion-
extension task and can unload the human joint from undesired
translational forces.
C. Passive-compliance and torque control modalities
The experimental characterization of the passive-compliance
control and the torque control proved the usability of
NEUROExos in different rehabilitation therapies.
Considering both control modes, the joint stiffness can be
tuned from near-zero to about 60 N·m/rad (see Fig. 9 and Fig.
10). The robot is therefore suitable to execute both robot-in-
charge and patient-in-charge exercises and, as a consequence,
to assist the movement of users with different level of
impairment.
Passive-compliance control) Results of the static
characterization (Fig. 9-a and Table I) prove that, by
regulating the resting length of the two springs through the
command, the passive compliance of NEUROExos can be tuned from 24 N m/rad up to 57 N m/rad, with increasing from 0 to 4 millimeters.
The friction loss given by Bowden cables is relatively low
and does not affect significantly the passive elastic behaviour
of the joint. This is evidenced by two factors. First, static
friction torque (i.e. torque offset in Fig. 9-a) is relatively
small: it is 0.5 N m, equal to 3.5% of the maximum torque output. Second, the sitffness range is close to the nominal
values calculated through (6) ( N m/rad), 5-8% lower than the measured. Furthermore, the passive stiffness is highly linear, as shown by the maximum
RMSE of the fitting in Fig. 10-a (see Table I), spanning from
0.8% to 2.6% of the maximum torque (i.e. 5 N m). The results of the dynamic characterization (Fig. 9-b, c, d
and Table II-V) show that the passive-compliance control can
be used to displace the user’s elbow, along a desired trajectory
with different levels of stiffness, full stable dynamic behaviour
and adequate bandwidth.
Bode diagram of Fig. 9-c shows that has a resonance frequency around 4.5-5 Hz. Since most rehabilitation tasks are
commonly limited to about 1 Hz [11], [24], [25], [28], the
passive-compliance control will guarantee 0 dB and full
stability.
Dynamic responses also show that, by increasing the joint
stiffness, the system becomes faster (i.e., increase of natural
frequency, decrease of rise time) and more damped (i.e.,
decrease of overshoot and resonance peak). The increased
damping is likely due to the Bowden transmission combined
with the antagonistic actuation. By increasing the joint
stiffness, and thus the preloading force, we get a higher cable
friction and, consequenlty, higher joint viscosity and damping
[57], [64]. This behaviour enhances the safety of NEUROExos
in the stiffer range, with the system capable of better
absorbing the effect of undesired (e.g. spastic) movements of
the user.
Results of the prototypical task, shown in Fig. 9-d,
demonstrate that NEUROExos can drive the human elbow
along a desired path in a stable and compliant manner. The
passive-compliance control softly attracts the human elbow on
the desired path and allows a difference between and
( ), which lowers with increasing. This feature is important to successful restore the patient’s motor function.
An adjustable soft assistance allows the adaptation of the
torque field to the specific level of impairment of the patient
and promotes a gradual active involvement of the subject [11],
[74]-[77].
Steady-state is significantly higher when driving the arm (Fig. 9-d) than when empty (Table III and Table V). This
discrepancy is related to the compliant behaviour of the
control and actuation system, which allows a which depends upon the stiffness coefficient and the loading
conditions. When the arm is fit in the exoskeleton,
gravitational ( kg) and inertial ( m) loads are much higher than in the free movement conditions, and thus
is also higher. Torque control) By using the same actuation system we
could implement a closed-loop torque control. The dynamic
characterization of the controller showed a -3dB bandwidth of
10.35 Hz which is sufficiently high for most rehabilitation
exercises, given the fact that the human arm can produce
muscular torque with a bandwidth of about 3.5 Hz [78].
Importantly, the system shows low output impedance over
the typical bandwidth of the human arm movement. This
means that if users can actively move their arm, the
exoskeleton would have a minimal load on it avoiding to
increase the effort. Under the action of the torque control, the
joint output impedance is lowered (by 29 dB) to 1 N·m/rad
and (by 9 dB) to 10 N·m/rad, during respectively 0.3 Hz and
3.2 Hz motion, compared to the passive impedance of 30
N·m/rad for (see (15)). Over the band 0.3-3.2 Hz, the measured values of parasitic
torque (and stiffness) are relatively low and comparable to the
ones reported in state-of-the-art robots [48]. Furthermore, over
Paper 12-0220
13
the typical frequency spectrum of rehabilitation tasks (which
is usually upper limited to about 0.5-1 Hz), NEUROExos
stiffness is as low as 1.5 N m/rad, introducing parasitic torque peaks (see also Fig. 10-c, d) which are negligible during the
execution of active movements, as demonstrated by the
experiments carried out in [38]-[40]. Indeed, as shown in [38],
wide bandwidth and minimum parasitic stiffness allow the
torque control to be used in a hierarchical control strategy,
where a higher control layer sets the desired value of according to a defined assistance strategy which partly
compensates for inertia, viscosity and gravity torque of the
elbow-NEUROExos coupled system.
Actuation and control hardware modules) The hardware
modules of NEUROExos actuation and control system have
been designed in order to fit a typical clinical environment for
physical rehabilitation. The remote position of the actuation
system allows to reduce the encumbrance and mass of the part
of the robot that is actually worn by the user. This increase the
performance of the systems as well as its acceptability.
V. CONCLUSION
In this paper, we presented NEUROExos, a novel powered
exoskeleton for elbow rehabilitation. NEUROExos possesses
three main innovative features: the double-shelled links, the
four DOF passive mechanism and a compliant antagonistic
actuation system. These features address three important
design requirements for a dependable device for physical
rehabilitation: (1) a wide and comfortable human-robot
physical interface which can gently transmit the interaction
torque, (2) the kinematic compatibility between the human and
the exoskeleton, to ensure a proper torque transmission to the
human joint without the risk of overloading the patient's
articulations, and (3) a safe and effective actuation system,
which can allow the execution of both robot-in-charge and
patient-in-charge rehabilitation exercises. In this paper, the
design and development of the system was described in detail,
in assocition with experimental characterization performed to
assess its effectiveness in a working scenario.
Future works will aim at using NEUROExos to carry out
post-stroke rehabilitation trials inside a clinica setting.
Attention will be also devoted to explore design solutions for
developing a more compact acutation and control system,
based on the use of electromagnetic motors and embedded
control units.
REFERENCES
[1] The world health report 2008, http://www.who.int/whr/2008/en [2] T. Ingall, “Stroke-Incidence, mortality, morbidity and risk”, Journal of
Insurance Medicine, vol. 36(2), pp. 143-152, 2004.
[3] V. L. Roger et al., “Heart Disease and Stroke Statistics—2011 Update: A Report From the American Heart Association”, Circulation 123: e18-209e; published online before print as
doi:10.1161/CIR.0b013e3182009701
[4] On the situation of women with disabilities in the European Union, http://www.europarl.europarl.europa.eu/ ,2007.
[5] Healthy ageing: keystone for a sustainable Europe, http://ec.europa.eu/health7ph_information/indicators /docs, 2007.
[6] G. Prange, M. Jannink, C. Groothuis-Oudshoorn, H. Hermens, and M. IJzerman, “Systematic review of the effect of robot-aided therapy on
recovery of the hemiparetic arm after stroke”, Journal of Rehabilitation
Research and Development, vol. 43(2), pp. 171-184, 2006. [7] S. Barreca, S.L. Wolf, S. Fasoli, and R. Bohannon, “Treatment
interventions for the paretic upper limb of stroke survivors: a critical
review”, Neurorehabilitation and Neural Repair, vol. 17(4), pp. 220-226, 2003.
[8] H.M. Feys, W.J. De Weerdt, B.E. Selz, G.A. Cox Steck, R. Spichiger, C.E. Vereeck, K.D. Putman, G.A. Van Hoydonock, “Effect of a therapeutic intervention for the hemiplegic upper limb in the acute phase
after stroke. A single-blind, randomized, controlled multicenter trial”,
Stroke, vol. 29(4), pp. 785-792, 1999. [9] G. Kwakkel, R.C. Wagenaar, J.W. Twisk , G.J. Lankhorst, J.C. Koetsier,
“Intensity of leg and arm training after primary middle-cerebral-artery
stroke: a randomized trial”, Lancet, vol. 354(9174), pp.191-196, 1999. [10] G. Kwakkel, B. Kollen, and H. Krebs, “Effects of robot-assisted therapy
on upper limb recovery after stroke: A systematic review”,
Neurorehabilitation and Neural Repair, vol. 22, pp.111-121, 2007. [11] H.I. Krebs, N. Hogan, M.L. Aisen, and B.T. Volpe, “Robot-Aided
Neurorehabilitation”, IEEE Transactions on Rehabilitation Engineering
vol. 6(1), pp. 75-87, 1998. [12] A.C. Lo, P.D. Guarino, L.G. Richards, J.K. Haselkorn, G.F. Wittenberg,
D.G. Federman, R.J. Ringer, T.H. Wagner, H.I. Krebs, B.T. Volpe, C.T.
Bever, D.M. Bravata, P.W. Duncan, B.H. Corn, A.D.Maffucci, S.E. Nadeau, S.S. Conroy, J.M. Powell, G.D. Huang, and P. Peduzzi, “Robot-
Assisted Therapy for Long-Term Upper-Limb Impairment after Stroke”,
The New England Journal of Medicine, vol. 362, pp. 1772-1783, 2010. [13] S.E. Fasoli, H.I. Krebs, J.Stein, W.R. Frontera, and N. Hogan, “Effects
of Robotic Therapy on Motor Impairement and Recovery in Chronic Stroke”, Archives of Physical Medicine and Rehabilitation, vol. 84(4),
pp. 477-482, 2003.
[14] P.S. Lum, C.G. Burgar, D.E. Kenney, and H.F. Machiel Van der Loos, “Quantification of Force Abnormalities During Passive and Active-
Assisted Upper-Limb Reaching Movements in Post-Stroke
Hemiparesis”, IEEE Transactions on Biomedical Engineering, vol. 46(6), pp. 652-662, 1999.
[15] S. Micera, M.C. Carrozza, E. Guglielmelli, G. Cappiello, F. Zaccone, C. Freschi, R. Colombo, A. Mazzone, C. Del conte, F. Pisano, G. Minuto, and P. Dario, “A Simple Robotic System for Neurorehabilitation”,
Autonomous Robots, vol. 19(3), pp. 271-284, 2006.
[16] D.J. Reinkensmeyer, C.D. Takahashi, W.K. Timoszyk, A.N. Reinkensmeyer, and L.E. Kahn, “Design of robot assistance for arm
movement therapy following stroke”, Advanced Robotics, vol. 14(7), pp.
625-637, 2001. [17] R. Loureiro, F. Amirabdollahian, M. Topping, B. Dressen, and W.
Harwin, “Upper Limb Robot Mediated Stroke Therapy – GENTLE/s
Approach”, Autonomous Robot, vol.15(1), pp. 35-51, 2003. [18] D. Mayhew, B. Bachrach, W. Rymer, and R. Beer, “Development of the
MACARM-a Novel Cable Robot for upper Limb Neurorehabilitation”,
in Proc. of the IEEE International Conference on Rehabilitation
Robotics, Chicago, US, pp. 299-302, 2005.
[19] A. Stienen, E. Hekman, F. Van der Helm, G. Prange, M. Jannink, A. Aalsm, and H. Van der Kooij, “Freebal: Dedicated gravity compensation for the upper extremities“, in Proc. of the International Conference on
Rehabilitation Robotics, Noordwijk, The Netherlands, pp. 804-808,
2007. [20] Z. Jia-Fan, Y. Can.Jun, C. Ying, Z. Yu, and D. Yi-Ming, “Modeling and
control of a curved pneumatic muscle actuator for wearable elbow
exoskeleton”, Mechatronics, vol. 18, pp. 448-457, 2008. [21] K. Kiguchi, T. Tanaka, and T. Fukuda, “Neuro-fuzzy control of a robotic
exoskeleton with EMG signals”, IEEE Transactions on Fuzzy Systems,
vol. 12(4), pp. 481-490, 2004. [22] J.C. Perry, J. Rosen, and S. Burns, “Upper-Limb Powered Exoskeleton
Design”, IEEE/ASME Transactions on Mechatronics, vol. 12(4), pp.
408-417, 2007.
[23] A. Frisoli, L. Borelli, A. Montagner, S. Marcheschi, C. Procopio, F. Salsedo, M. Bergamasco, M.C. Carboncini, M. Tolaini, and B. Rossi,
“Arm rehabilitation with a robotic exoskeleton in Virtual Reality”, in Proc. of IEEE International Conference on Rehabilitation Robotics,
Noordwijk, The Netherlands, pp. 631-642, 2007.
[24] A. Frisoli, F. Salsedo, M. Bergamasco, B. Rossi, and M.C. Carboncini, “A force-feedback exoskeleton for upper-limb rehabilitation in virtual
reality”, Applied Bionics and Biomechanics, 1754-2103, vol. 6(2), pp.
115-126, 2009.
http://www.who.int/whr/2008/enhttp://www.europarl.europarl.europa.eu/http://ec.europa.eu/health7ph_information/indicators%20/docs
Paper 12-0220
14
[25] N.G. Tsagarakis and D.G. Caldwell, “Development and Control of a ‘Soft-Actuated’ Exoskeleton for Use in Physiotherapy and Training”,
Autonomous Robots, vol. 15, pp. 21-23, 2003.
[26] C. Carignan, J. Tang, S. Roderick, and M. Naylor, “A Configuration-Space Approach to Controlling a Rehabilitation Arm Exoskeleton”, in Proc. IEEE International Conference on Rehabilitation Robotics,
(ICORR), Noordwijk, The Netherlands, pp. 524-531, 2007.
[27] E. Rocon, J.M. Belda-Lois, A.F. Ruiz, M. Manto, J.C. Moreno, and J.L. Pons, “Design and Validation of a Rehabilitation Robotic Exoskeleton
for Tremor Assessment and Suppression”, IEEE Transactions on Neural
Systems and Rehabilitation Engineering, vol. 15(3), pp. 367-378, 2007. [28] T. Nef, M. Mihelj, and R. Riener, “ARMin: A robot for patient-
cooperative arm therapy”, Medical and Biological Engineering and
Computing, vol. 45(9), pp. 887-900, 2007. [29] M. Mihelj, T. Nef, and R. Riener, “ARMin II – 7 DoF rehabilitation:
mechanics and kinematics”, in Proc. of IEEE International Conference
on Robotics and Automation (ICRA), pp. 4120-4125, 2007. [30] R.J. Sanchez, E. Wolbrecht, R. Smith, J. Liu, S. Rao, S. Cramer, T.
Rahman, J.E. Bobrow, and D.J. Reinkensmeyer, “A Pneumatic Robot
for Re-Training Arm Movement after Stroke: Rationale and Mechanical Design”, in Proc. of the IEEE International Conference on
Rehabilitation Robotics (ICORR), pp. 500-504, 2005.
[31] R.J. Sanchez, J. Liu, S. Rao, P. Shah, R. Smith, T. Rahman, S.C. Cramer, J.E. Bobrow, and D.J. Reinkensmeyer, “Automating Arm
Movement Training Following Severe Stroke: Functional Exercises
With Quantitative Feedback in a Gravity-Reduced Environment”, IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol.
14(3), pp. 378-389, 2006. [32] A. Schiele and F.C. van der Helm, “Kinematic Design to Improve
Ergonomics in Human Machine Interaction”, IEEE Transaction on
Neural Systems and Rehabilitation Engineering, vol. 14(4), pp. 456–469, 2006.
[33] A.M. Dollar and H. Herr, “Lower Extremity Exoskeletons and Active Orthoses: Challenges and State-of-the-Art”, IEEE Transactions on Robotics, vol. 24, pp. 144-158, 2008.
[34] A.H.A. Stienen, E.E.G. Hekman, F.C.T. van der Helm, and H. van der Kooij, “Self-Aligning Exoskeleton Axes Trough Decoupling of Joint Rotations and Translations, IEEE Transactions on Robotics, vol. 25(3),
pp. 628-633, 2009.
[35] J.L. Pons, “Rehabilitation Exoskeletal Robotics”, IEEE Engineering in Medicine and Biology Magazine, vol. 29(3), pp. 57-63, 2010.
[36] T. Lenzi, S.M.M. De Rossi, N. Vitiello, A. Chiri, S. Roccella, F. Giovacchini, F. Vecchi, and M.C. Carrozza, “The neuro-robotics paradigm: NEURARM, NEUROExos, HANDEXOS”, in Proc. of the
International Conference of the IEEE Engineering in Medicine and
Biology Society, Minneapolis, US, pp. 2430-2433, 2009. [37] T. Lenzi, N. Vitiello, S.M.M. De Rossi, S. Roccella, F. Vecchi, and
M.C. Carrozza, “NEUROExos: a variable impedance powered elbow
exoskeleton”, in Proc. of the IEEE International Conference on Robotics
and Automation (ICRA),Shanghai, China, pp.1419-1426, 2011.
[38] T. Lenzi, S.M.M. De Rossi, N. Vitiello, and M.C. Carrozza, “Proportional EMG control for upper-limb powered exoskeletons”, in Proc. of the International Conference of the IEEE Engineering in
Medicine and Biology Society, Boston, US, pp. 628-631, 2011.
[39] T. Lenzi, S.M.M. De Rossi, N. Vitiello, and M.C.C. Carrozza, “Intention-Based EMG Control for Powered Exoskeletons”, IEEE
Transactions on Biomedical Engineering, vol. 59(8), pp. 2180- 2190,
2012. [40] R. Ronsse, N. Vitiello, T. Lenzi, J. Van den Kieboom, M.C. Carrozza,
and A.J. Ijspeert, “Human-robot synchrony: flexible assistance using
adaptive oscillators”, IEEE Transactions on Biomedical Engineering, vol. 58(4), pp. 1001-1012, 2011.
[41] T. Lenzi, N. Vitiello, S.M.M. De Rossi, A. Persichetti, F. Giovacchini, S. Roccella, F. Vecchi, M.C. Carrozza, "Measuring Human-Robot
Interaction on Wearable Robots: a Distributed approach", Mechatronics,
vol. 21(6), pp. 1123-1131, 2011.
[42] A. Schiele, “Fundamental of Ergonomic Exoskeleton Robots”, Ph.D. dissertation, Technishe Universiteit Delft, 2008.
[43] A. Schiele and F.C.T van der Helm, “Kinematic Design to improve Ergonomics in Human-Machine Interaction”, IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 14(2), pp. 456-469,
2006.
[44] M. Bottlang, S.M. Madey, C.M. Steyers, J.L. Marsh, and T.D. Brown, “Factors influencing accuracy of screw displacement axis detection with
a DC-based electromagnetic tracking system”, Journal of Biomechanical
Engineering, vol. 120(3), pp. 431-435, 1998. [45] M. Bottlang, S.M. Madey, C.M. Steyers, J.L. Marsh, and T.D. Brown,
“Assessment of elbow joint kinematics in passive motion by
electromagnetic motion tracking”, Journal of Orthopaedic Research, vol. 18, pp. 195-201, 2000.
[46] T. R. Duck, C.E. Dunning, G.J.W. King, and J.A. Johnson, “Variability and repeatability of the flexion axis at the ulnohumeral joint”, Journal of Orthopaedic Research, vol. 21, pp. 399–404, 2003.
[47] A.H.A. Stienen, E.E.G. Hekman, H. ter Braak, A.M.M. Aalsma, F. van der Helm, and H. van der Kooij, “Design of a rotational hydro-elastic actuator for a powered exoskeleton for upper-limb rehabilitation”, IEEE
Transactions on Biomedical Engineering, vol. 57(3):728-735, 2010.
[48] J.F. Veneman, R. Ekkelenkamp, R. Kruidhof, F.C.T. van der Helm, and H. van der Kooij, “A Series Elastic- and Bowden-Cable-Based
Actuation of Use Torque Actuator in Exoskeleton-Type Robots”, The
International Journal of Robotics Research, vol. 25(3), pp. 261-281, 2006.
[49] M. Zinn, O. Khatib, B. Roth, and J.K. Salisbury, “A New Actuation Approach for Human-Friendly Robot Design”, The International Journal of Robotics Research, vol. 23(4-5), pp.379-398, 2004.
[50] G. Pratt and M.M. Williamson, “Series elastic actuators”, in Proc. of the IEEE International Conference on Intelligent Robots and Systems, Pittsburgh, US, pp. 339-406.
[51] J. Oblak, I. Cikajlo, and Z. Matjačić, “Universal Haptic Drive: A Robot for Arm and Wrist Rehabilitation”, IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 18(3), pp. 293-302, 2010.
[52] R. Schiavi, G. Grioli, S. Sen, and A. Bicchi, “VSA-II: a novel prototype of variable stiffness actuator for safe and performing robots interacting
with humans”, in Proc. of the IEEE International Conference on
Robotics and Automation, pp. 2171-2176, 2008. [53] S.A. Migliore, E.A. Brown, and S.P. De Weerth, “Biologically inspired
joint stiffness control”, in Proc. of the IEEE International Conference on
Robotics and Automation, pp. 4508-4513, 2005. [54] K. Koganezawa, T. Inaba, and T. Nakazawa, “Stiffness and angle
control of antagonistically driven joint”, in Proc. of the IEEE
International Conference on Biomedical Robotics and Biomechatronics, pp. 1007-1013, 2006.
[55] D.G. Caldwell, G.A. Medrano-Cerda, and M. Goodwin, “Control of pneumatic muscles actuators”, IEEE Control System Magazine, vol. 15(1), pp. 40-48, 1995.
[56] N. Vitiello, T. Lenzi, S.M.M. De Rossi, S. Roccella, M.C. Carrozza, “A sensorless torque control for Antagonistic Driven Compliant Joints”, Mechatronics, vol. 20(3), pp. 355-367, 2010.
[57] T. Lenzi, N. Vitiello, J. McIntyre, S. Roccella, M.C. Carrozza, “A robotic model to investigate the human motor control”, Biological Cybernetics, vol. 105(1), pp. 1-19, 2011.
[58] E. Cattin, S. Roccella, N. Vitiello, I. Sardellitti, K.A. Panagiotis, P. Vacalebri, F. Vecchi, M.C. Carrozza, K. Kyriakopulos, and P. Dario,
“Design and Development of a Novel Robotic Platform for Neuro-
Robotics Applications: the NEURobotics ARM (NEURARM)”,
Advanced Robotics, Special Issue on Robotics platforms for Neuroscience, vol. 22, pp. 3-37, 2008.
[59] N. Vitiello, E. Cattin, S. Roccella, F. Giovacchini, F. Vecchi, M.C. Carrozza, and P. Dario, “The NEURARM towards a Platform for Joint Neuroscience Experiments on Human Motion Control Theories”, in
Proc. IEEE International Conference on Intelligent Robots and Systems,
San Diego, pp. 1852-1857, 2007. [60] N. Vitiello, T. Lenzi, J. McIntyre, S. Roccella, E. Cattin, F. Vecchi, and
M.C. Carrozza, “Characterization of the NEURARM bio-inspired joint
position and stiffness open loop controller”, in Proc. of IEEE International Conference on Biomedical Robotics and Biomechatronics,
Scottsdale, US, pp. 138-143, 2008.
[61] M.O. Abe and N. Yamada, “Modulation of elbow joint stiffness in a vertical plane during cyclic movement at lower or higher frequencies
than natural frequency”, Experimental Brain Research, vol. 153, pp.
394-299, 2003. [62] D.W. Franklin, E. Burdet, R. Osu, M. Kawato, and T.E. Milner,
“Functional significance of stiffness in adaptation of multijoint arm
movements to stable and unstable dynamics, Experimental Brain Research, vol. 151, pp. 145-157, 2003.
[63] A. Schiele, P. Letier, R. van der Linde, and F.C.T. van der Helm, “Bowden cable actuator for force-feedback exoskeletons”, in Proc. of IEEE International Conference on Intelligent Robots and Systems, pp.
3599-3604, 2006.
Paper 12-0220
15
[64] A. Chiri, N. Vitiello, F. Giovacchini, S. Roccella, F. Vecchi, and M.C. Carrozza, “Mechatronic Design and Characterization of the Index Finger
Module of a Hand Exoskeleton for Post-stroke Rehabilitation”,
IEEE/ASME Transactions on Mechatronics, in press, doi:
10.1109/TMECH.2011.2144614, 2011. [65] N. Hogan, “An organizing principle for a class of voluntary
movements”, Journal of Neuroscience, vol. 4, pp. 2745-2754, 1984.
[66] N. Hogan, E. Bizzi, and F.A. Mussa-Ivaldi, T. Flash, “Controlling multijoint motor behaviour”, Exercise and Sport Sciences Reviews, vol.
15, pp. 153-190, 1987.
[67] E. Bizzi, N. Accornero, W. Chapple, and N. Hogan, “Posture control and trajectory formation during arm movement”, Journal of Neuroscience,
vol. 4, pp. 2738-2744, 1984.
[68] A. Polit and E. Bizzi, “Characteristics of motor programs underlying arm movements in monkeys”, Journal of Neurophysiology, vol. 42, pp.
183-194, 1979.
[69] A.G. Feldman, “Functional tuning of the nervous system with control of movement or maintenance of steady posture, II: controllable parameters
of the muscles”, Biophysics, vol. 11, pp. 565-578, 1966.
[70] A. Bicchi and G. Tonietti, “Fast and Soft Arm Tactics: Dealing with the Safety-Performance Trade-Off in Robot Arms Design and Control”,
IEEE Robotics and Automation Magazine, vol. 11(2), pp. 22-33, 2004.
[71] I.A. Kapandji, “Fisiologia articolare. Arto Superiore”, Editions MALOINE, Paris, France, 2002.
[72] J.L. Pons, “Wearable Robots: Biomechatronic Exoskeletons”, Wiley-Blackwell, Hoboken (2008).
[73] S.M.M. De Rossi, N. Vitiello, T. Lenzi, R. Ronsse, B. Koopman, A. Persichetti, F. Vecchi, A.J. Ijspeert, H. van der Kooij, and M.C. Carrozza, “Sensing Pressure Distribution on a Lower-Limb Exoskeleton
Physical Human-Machine Interface”, Sensors, vol. 11(1), pp.: 207-227,
2011. [74] M. Ziegler, H. Zhong, R.R. Roy, and V. Reggie Edgerton, “Why
Variability Facilitates Spinal Learning”, Journal of Neuroscience, vol.
30(32), pp. 10720-10726, 2010. [75] L.L. Cai, A.J. Fong, C.K. Otoshi, Y. Liang, J.W. Burdick, R.R. Roy, and
V. Reggie Edgerton, “Implications of Assist-As-Needed Robotic Step
Training after a Complete Spinal Cord Injury on Intrinsic Strategies of Motor Learning”, Journal of Neuroscience, vol. 26(41), pp. 10564-
10568, 2006.
[76] J.L. Patton, M.E. Stoykov, M. Kovic, and F.A. Mussa-Ivaldi, “Evaluation of robotic training forces that either enhance or reduce error
in chronic hemiparetic stroke survivors”, Experimental Brain Research,
vol. 168, pp. 368-383, 2006. [77] H.I. Krebs, J.J. Palazzolo, L. Dipietro, M. Ferraro, J. Krol, K.
Rannekleiv, B.T. Volpe and N. Hogan, “Rehabilitation Robotics:
Performance-Based Progressive Robot-Assisted Therapy Autonomous Robots”, Autonomous Robot, vol. 15(1), pp. 7-20, 2003.
[78] H.S. Milner-Brown, R.B. Stein, and R. Yemm, “The contractile properties of human motor units during voluntary isometric
contractions”, Journal of Physiology, vol. 228, pp.: 285-306, 1973.
Nicola Vitiello (M’12) is Assistant Professor at The BioRobotics Institute of Scuola Superiore Sant’Anna. He
received the M.Sc. degree in Biomedical Engineering (cum
laude) both from the University of Pisa, Italy, in 2006, and from Scuola Superiore Sant’Anna, Italy, in 2007. He
received the Ph.D. degree in Biorobotics from the Scuola
Superiore Sant'Anna, Italy, in 2010. His main research interests include the development of wearable robotic devices for human
motion assistance and rehabilitation and of robotic platforms for
neuroscientific investigations. He is author and co-author of 11 ISI papers and 23 peer-review conference-proceedings papers. He served as scientific
secretary of the EU FP7 CA-RoboCom project, and he is currently project
manager of the EU FP7 CYBERLEGs Project.
Tommaso Lenzi (S’10) received the B.Sc and M.Sc in
Biomedical Engineering (cum laude) from University of Pisa in 2005 and 2008 respectively. Since 2008 is a PhD
student in BioRobotics at The BioRobotics Institute
(former ARTS Lab) of Scuola Superiore Sant'Anna (Italy). From August 2011 to February 2012 he has been a Visiting
Research Scientist at the Mechanical Engineering
Department of University of Delaware (USA) under the supervision of Dr. S. K. Agrawal. His main research interests lies in between
robotics, biomechanics and neurophysiology with a major in design and
control of powered exoskeletons for human assistance and rehabilitation. He
is author and co-author of 8 ISI Journal papers and 16 peer-review
conference-proceedings papers and serves as a reviewer of IEEE/ASME
Transaction on Mechatronics, and Applied Bionics and Biomechanics. From
2010 Mr. Lenzi is student member of IEEE, RAS and EMBS scientific society.
Stefano Roccella received his Laurea Degree in Aeronautical Engineering from the University of Pisa. In
1999 he joined The BioRobotics Institute of Scuola
Superiore Sant’Anna as a Research Assistant. In March 2006, he received his Ph.D. in Robotics at University of
Genova. Currently, he is a PhD Technical Research
Manager at Scuola Superiore Sant’Anna. His research interests are in the fields of mechatronic systems, humanoid robotics,
biomedical robotics, rehabilitation engineering, biomechatronics, and MEMS
design and development.
Stefano Marco Maria De Rossi (S'10) received the M.Sc.
degree in control engineering (cum laude) both from the University of Pisa, Italy and from Scuola Superiore
Sant'Anna, Pisa, Italy in 2009. He is currently working
toward the Ph.D. degree in Biorobotics at The BioRobotics Institute, Scuola Superiore Sant’Anna, Pisa, Italy. His
current research interests include wearable robotics,
human-robot interfaces and robotics for upper and lower limb rehabilitation.
Emanuele Cattin received his Laurea Degree in Mechanical Engineerig from the University of Padova in
July 2003, and the Ph.D. degree in Robotics from the
University of Genova in 2008. In January 2004 he joined the Advanced Robotic Technology and Systems (ARTS)
Lab (now The BioRobotics Institute) at the Scuola
Superiore Sant’Anna in Pisa, Italy., where he conducted research activities in the field of wearable robotics until 2009.
Francesco Giovacchini received his M.S. degree in Mechanical Engineering in 2006. Since January 2007 he
has worked at The BioRobotics Institute of Scuola
Superiore Sant’Anna. His research interests are in the field of biomechatronics, and more specifically in the
development of robotic platforms and wearable devices for
rehabilitation and neurobotics.
Fabrizio Vecchi (M’04) received the University Degree (Laurea) in electrical engineering from the University of
Pisa, Italy, in 1999 and the Ph.D. in Robotics from the
University of Genova in 2003. He is Post-Doc research
scientist at The BioRobotics Institute of Scuola Superiore
Sant’Anna. During his career he participated in the design
and development of human-machine interfaces, biomechatronic prosthetic and robotic hands, passive and
active orthoses, flexible sensory skin, and technical assistive devices. He is
co-inventor of 7 national and international patents, and is co-author of more than 40 scientific papers published on ISI journals and on proceedings of
international conferences.
Maria Chiara Carrozza (Master in Physics at University
of Pisa in 1990, PhD in Engineering at Scuola Superiore
Sant'Anna in 1994) is Full Professor of Biomedical Engineering and Robotics at Scuola Superiore Sant’Anna
(SSSA). Since Nov. 2007, she is Rector of Scuola
Superiore Sant’Anna. Her research interests are in ambient
assisted living, biorobotics, rehabilitation engineering,
bionics, cybernetic hands, humanoid robotics, systems for
functional replacements and augmentation, biomechatronic interfaces, tactile sensors and artificial skin. She coordinates several research projects funded by
national and European grants in the area of rehabilitation engineering,
prosthetics, human-machine interfaces and exoskeleton for functional replacement and assistance. She is author of several scientific papers (more
than 60 ISI papers and more than 100 papers in referred conference
proceedings) and of 12 national and international patents. She is Member of IEEE Robotics and Automation and Engineering in Medicine and Biology
Societies.